Sunteți pe pagina 1din 9

# Section 1.

14

## 1.14 Tensor Calculus I: Tensor Fields

In this section, the concepts from the calculus of vectors are generalised to the calculus of
higher-order tensors.

1.14.1

Tensor-valued Functions

## Tensor-valued functions of a scalar

The most basic type of calculus is that of tensor-valued functions of a scalar, for example
the time-dependent stress at a point, S S(t ) . If a tensor T depends on a scalar t, then
the derivative is defined in the usual way,
dT
T(t t ) T(t )
,
lim t 0
dt
t

## which turns out to be

dT dTij

ei e j
dt
dt

(1.14.1)

The derivative is also a tensor and the usual rules of differentiation apply,
d
T B dT dB
dt
dt
dt
d
(t )T dT d T
dt
dt
dt
d
Ta T da dT a
dt
dt dt
d
TB T dB dT B
dt
dt
dt

d T dT
T

dt
dt

## For example, consider the time derivative of QQ T , where Q is orthogonal. By the

product rule, using QQ T I ,

d
dQ T
dQ T dQ T
dQ
QQ T
Q Q
Q Q

0
dt
dt
dt
dt
dt
Thus, using Eqn. 1.10.3e

Q T Q Q
T Q
QT
Q

## Solid Mechanics Part III

115

(1.14.2)

Kelly

Section 1.14

Q T is a skew-symmetric tensor.
which shows that Q

1.14.2

Vector Fields

The gradient of a scalar field and the divergence and curl of vector fields have been seen
in 1.6. Other important quantities are the gradient of vectors and higher order tensors
and the divergence of higher order tensors. First, the gradient of a vector field is
introduced.
The Gradient of a Vector Field
The gradient of a vector field is defined to be the second-order tensor

a
a
e j i ei e j
x j
x j

## Gradient of a Vector Field

(1.14.3)

In matrix notation,
a1

x1
a
x
a1
3
x1

a1
x 2
a 2
x 2
a3
x 2

a1

x3
a 2
x3
a3
x3

(1.14.4)

## One then has

ai
e i e j dxk e k
x j
ai
dx j e i
x j

(1.14.5)

da
a(x dx) a(dx)
which is analogous to Eqn 1.6.10 for the gradient of a scalar field. As with the gradient
of a scalar field, if one writes dx as dx e , where e is a unit vector, then

da
dx in e direction

(1.14.6)

Thus the gradient of a vector field a is a second-order tensor which transforms a unit
vector into a vector describing the gradient of a in that direction.

116

Kelly

Section 1.14

## As an example, consider a space curve parameterised by s, with unit tangent vector

dx / ds (see 1.6.2); one has

da a dx j a

e j

x j

ds x j ds x j

Although for a scalar field grad is equivalent to , note that the gradient defined in
1.14.3 is not the same as a . In fact,

a T

(1.14.7)

since
a ei

a j

a je j
ei e j
xi
xi

(1.14.8)

## a j / xi e i e j , are both commonly used. In what follows, they will be distinguished by

labeling the former as grada (which will be called the gradient of a) and the latter as
a.
Note the following:
in much of the literature, a is written in the contracted form a , but the more
explicit version is used here.
some authors define the operation of on a vector or tensor not as in 1.14.8, but
through / xi e i so that a grada ai / x j ei e j .

## Consider a particle p 0 of a deforming body at position X (a vector) and a neighbouring

point q 0 at position dX relative to p 0 , Fig. 1.14.1. As the material deforms, these two
particles undergo displacements of, respectively, u(X) and u ( X dX) . The final
positions of the particles are p f and q f . Then

dx dX u( X dX) u( X)
dX du( X)

117

Kelly

Section 1.14

initial

q0

u ( X dX )

dX

p0

qf
u ( X)

dx
pf

final

## Figure 1.14.1: displacement of material particles

Thus the gradient of the displacement field u encompasses the mapping of (infinitesimal)
line elements in the undeformed body into line elements in the deformed body. For
example, suppose that u1 kX 22 , u 2 u 3 0 . Then

0 2kX 2
u i
0
0
X j
0
0

0
0 2kX 2 e1 e 2
0

## A line element dX dX i e i at X X i e i maps onto

dx dX 2kX 2 e1 e 2 dX 1e1 dX 2 e 2 dX 3 e 3
dX 2kX 2 dX 2 e1
The deformation of a box is as shown in Fig. 1.14.2. For example, the vector dX de 2
(defining the left-hand side of the box) maps onto dx 2k de1 e 2 .

X2

final

X1
Figure 1.14.2: deformation of a box
Note that the map dX dx does not specify where in space the line element moves to.
It translates too according to x X u .

## The Divergence and Curl of a Vector Field

The divergence and curl of vectors have been defined in 1.6.6, 1.6.8. Now that the
gradient of a vector has been introduced, one can re-define the divergence of a vector
independent of any coordinate system: it is the scalar field given by the trace of the

118

Kelly

Section 1.14

## Divergence of a Vector Field (1.14.9)

Similarly, the curl of a can be defined to be the vector field given by twice the axial
vector of the antisymmetric part of grad a .

1.14.3

Tensor Fields

A tensor-valued function of the position vector is called a tensor field, Tijk (x) .
The Gradient of a Tensor Field

The gradient of a second order tensor field T is defined in a manner analogous to that of
the gradient of a vector, Eqn. 1.14.2. It is the third-order tensor

Tij
T
ek
ei e j e k
x k
x k

T ei

T jk

ei e j ek
T jk e j e k
xi
xi

(1.14.11)

## The Divergence of a Tensor Field

Analogous to the definition 1.14.9, the divergence of a second order tensor T is defined to
be the vector
div T grad T : I

Tij
x j

(T jk e j e k )
T
ei
ei
xi
xi

## Divergence of a Tensor (1.14.12)

ei

The divergence of a tensor can also be equivalently defined as that vector field which
satisfies the relation

divT v divT T v
for all constant vectors v.
One also has
T ei

## Solid Mechanics Part III

T ji

(T jk e j e k )
ei
xi
x j

119

(1.14.13)

Kelly

Section 1.14

so that
divT T T

(1.14.14)

As with the gradient of a vector, both Tij / x j e i and T ji / x j e i are commonly used
as definitions of the divergence of a tensor. They are distinguished here by labelling the
former as divT (called here the divergence of T) and the latter as T . Note that the
operations divT and T are equivalent for the case of T symmetric.
Note the following
some authors define the operation of on a vector or tensor not as in (1.14.13),
but through / xi ei so that T divT Tij / x j ei .

using the convention that the dot is omitted in the contraction of tensors, one should
write T for T , but the dot is retained here because of the familiarity of this latter
notation from vector calculus.
another operator is the Hessian, 2 / xi x j e i e j .

Identities

Here are some important identities involving the gradient, divergence and curl
{Problem 5}:
div u v grad u v (div v )u
T

(1.14.15)

divAv v divA T tr Agradv

(1.11.16)

Note also the following identities, which involve the Laplacian of both vectors and
scalars:
2 u v 2 u v 2gradu : gradv u 2 v
curl curl u graddiv u 2 u

120

(1.14.17)

Kelly

Section 1.14

1.14.4

## Cylindrical and Spherical Coordinates

Cylindrical and spherical coordinates were introduced in 1.6.10 and the gradient and
Laplacian of a scalar field and the divergence and curl of vector fields were derived in
terms of these coordinates. The calculus of higher order tensors can also be cast in terms
of these coordinates.
For example, from 1.6.30, the gradient of a vector in cylindrical coordinates is
T

1

e z u r e r u e u z e z
e
r
z
r

u
u
1 u r u
r er er
e r e r e r e z
r
r
z
r

u
u
1 u u r
e e r
e e e e z
r
z
r
r
u
u
1 u z
z ez er
e z e z e z e z
z
r
r

(1.14.18)

{Problem 6}
1 Ar Arz Arr A
A

divA A T rr
e r
z
r
r
r

1 A Az Ar Ar
A

r
e
z
r
r

(1.14.19)

A
1 Az Azz
A

zr zr
e z
z
r
r
r

1.14.5

## The Divergence Theorem

The divergence theorem 1.7.12 can be extended to the case of higher-order tensors.
Consider an arbitrary differentiable tensor field Tijk ( x, t ) defined in some finite region of
physical space. Let S be a closed surface bounding a volume V in this space, and let the
outward normal to S be n. The divergence theorem of Gauss then states that

ijk

nk dS
V

Tijk
x k

(1.14.20)

dV

Tn dS div T dV ,
S

121

Tij

T n dS x
ij

dV

(1.14.21)

Kelly

Section 1.14

TndS div ( T) dV
S

S

u TndS div (T
S

1.14.6

(1.14.22)

u) dV

## Formal Treatment of Tensor Calculus

Following on from 1.6.12, here a more formal treatment of the tensor calculus of fields
is briefly presented.

## What follows is completely analogous to Eqns. 1.6.46-49.

A vector field v : E 3 V is differentiable at a point x E 3 if there exists a second
order tensor Dv x E such that
vx h v x Dv x h o h

for all h E

(1.14.23)

In that case, the tensor Dv x is called the derivative (or gradient) of v at x (and is given
the symbol vx ).
Setting h w in 1.14.23, where w E is a unit vector, dividing through by and
taking the limit as 0 , one has the equivalent statement
vx w

d
d

v x w for all w E

(1.14.24)

Using the chain rule as in 1.6.11, Eqn. 1.14.24 can be expressed in terms of the
Cartesian basis e i ,
vx w

vi
v
wk e i i e i e j wk e k
x k
x j

(1.14.25)

vx

vi
ei e j
x j

(1.14.26)

## which is Eqn. 1.14.3.

Solid Mechanics Part III

122

Kelly

Section 1.14

1.14.7

Problems

1. Consider the vector field v x12 e1 x32 e 2 x 22 e 3 . (a) find the matrix representation of
2. If u x1 x 2 x3e1 x1 x 2 e 2 x1e 3 , determine 2 u .
3. Suppose that the displacement field is given by u1 0, u 2 1, u 3 X 1 . By using
gradu , sketch a few (undeformed) line elements of material and their positions in the
deformed configuration.
4. Use the matrix form of gradu and u to show that the definitions
(ii) curla 2 , where is the axial vector of the skew part of grada
agree with the definitions 1.6.17, 1.6.21 given for Cartesian coordinates.
5. Prove the following:
(i)
(ii)
(iii)
(iv)
(v)

div u v grad u v (div v )u
T

(vi)